Deriving Inundation Risk Using Lidar and SLOSH Outputs

OK, so I have always wondered how to deal with the MOMs (maximum of MEOWs) from the SLOSH program since it really is a worst case scenario (Figure 1) for each hurricane category; and not likely going to come into play (thankfully). Then there are the MEOWs (maximum envelopes of water) for each direction and speed – which is quite a few options. They give you a glimpse of a ‘normal’ hurricane strike if you choose the correct speed and direction, which can be attained to some degree from historic hurricane tracks (see: http://coast.noaa.gov/hurricanes/). Then there are tides that also complicate things, which gets me back to my leading question, how to use this information to map present and future risk. The data is there, and it seems ready for consumption.

Figure 1. SLOSH MOM data for Category 1 hurricane at Logan Airport. Blue areas are under water at some level and note that nearly every part of the airport is flooded.

As I have written about previously, this kind of data is basically a normal population (sample) of the potential conditions at fairly precise geographic locations. To test the use of the data in mapping risk of inundation I chose to look at Logan Airport in Boston (since I was recently there). The collective data from all of the MEOWs, which themselves represents a couple of model runs, is akin to an ensemble of, in this case, category 1 hurricanes hitting the Boston Area.   

So I decided to use the z-score mapping technique, which required mapping and interpolating the data for each storm and generating a grid that had 36 different storm values across the Logan airport area. OK, yeah the data requires some interpolation since each storm’s data does not necessarily fill the whole grid. For example, surge values from many storms tracking easterly did not exist on the airport since there was no expectation of water to have flooded the land. The tact that I took was to imagine (model) the water levels as if the area was a huge pier, and water could run below it. This is not strictly correct, since the surge is modified by the land’s geometry, but for this example I had to be content with this assumption. Some inland grid cells did not have data from any storms and were ignored altogether, it was just the cells with data in some storms and not others that required some assumptions.

So, after generating a script to compute surge values for 36 storms and store them in a single file I had a 500 x 500 m grid (vector) of 36 different surge heights from the cat 1 storms. I choose to use VDatum to translate the values to mean higher high water; I could have used the MHHW values in SLOSH, but for this example I went with the VDatum conversion so I could include that error in the overall error. Using the slosh values, I derived the mean category 1 surge (without waves and rain) and the standard deviation of them to use for generating z-scores, which leads to percentages. The standard deviation of the lidar (RMSE) and the tidal correction was also included using the least squares method. The final standard deviations, or maximum cumulative error, were really driven by the storm surge variances, much less by the lidar or tidal correction. 

The final output is not a map of inundation, but rather the chance that, based on the data and assumptions, the location would be inundated (Figure 2). Comparing the inundation chances (Figure 2) to the MOM inundation surface (Figure 1) highlights the differences. In the worst case (MOM) the entire airport would essentially be under water to some degree; however, the inundation risk is really much lower for the airport. Most of airport has less than a 30% chance (i.e., 3 in 10 storms) of being inundated and many of the runways are even less. SLOSH outputs do not account for wave set-up or rain induced flooding, which are some things to consider, so the risks may be a bit conservative.

Figure 2. Map of inundation risk (chances) from a category 1 hurricane hitting the Boston Area at high tide. Only areas with a 20% or greater chance are shown. The red contour highlights areas with a 60% or higher chance of inundation from a category 1 storm.

The next question one may ask is “how would things look in the future”? I choose to look at a fairly small time horizon, and one that would potentially coincide with the future Olympic Games in Boston (taking place during hurricane season). I used the latest curves from NOAA to define the mean and the standard deviation of the SL estimates in 2025. Based on these curves the mean SL is about 13 cm higher than in 1992 (the middle of the present tidal epoch) with a standard deviation of 7.5 cm. The standard deviation was included in the MCU along with the standard deviation of the storms, lidar, and tidal surfaces. I could have included the local trends as well, but in this case they are only slightly different than the global trends and represent less than a couple cm difference in 2025.

Just as a point of comparison, in 2100 the mean value for global sea level rise is 98 cm with a standard deviation of 81 cm. In this case the standard deviation will significantly add to the level of uncertainty (MCU), as one would expect. In fact the standard deviation of the 2100 NOAA SLR predictions are about the same magnitude as the standard deviations for the SLOSH runs for a category 1 storm in Boston.     

That said, the 2025 results (Figure 3) are much more comparable to the present (i.e., 1992) conditions. There are slightly higher risks of inundation, noticeable with fewer runways open at the 20% risk level.

Figure 3.  2025 Category 1 storm inundation risks. The same color scheme and contour value was used on both Figures 2 and 3.

OK, so – very interesting, but how does this help? Well, here is what it means to me: for mapping potential future risks the MOM values are a bit over-alarming and the individual storms require too much foresight (prediction of a single future storm) whereas this technique, like the momma bear's bed, is a nice middle ground. It does not provide elevations of flooding (although the 50% contour is the discriminate value that is used in most mapping as – “the value”), but rather a choice of elevation values depending on your risk-tolerance. Another modification could include use of historic data to define the path of the most likely strike and look at several different variations on the theme based on the percentages (start to add strike percentages to the mix).

Bottom line, the data is there, and with some work can provide a more holistic view of surge flooding than when using single storms or MOM’s. Oh, getting back to Logan, I don’t know the actual on-the-ground risk level tolerance that an airport has, but I would assume that if a category 1 hurricane was predicted for the Boston area, Logan airport would want to take steps to protect infrastructure that had a 20% chance or higher level of risk. Which in this case is a significant amount of the airport infrastructure.

Hope this helps stimulate some ideas for you. And if this makes no sense or you have any questions about the technique please get in touch with me - keil@geosciconsultants.com - and we can discuss.